Advertisements

Archive

Archive for the ‘SQL 2008 R2’ Category

SQL Server 2012 NUMA Node Imbalance–Conclusion

October 4, 2015 Leave a comment

Final post in the series, I hope to summarize my findings for everyone.  Thanks for all the comments, feedback and comments I have received to-date on these posts.

Before providing the summary of all the testing, research and reading, I want to provide some internals for folks.

NUAM Memory Allocation

If a server has NUMA configuration, when we define the max memory setting.  All memory gets allocated evenly from all NUMA nodes.  For example, if I allocate 8GB from a server that has 4 NUMA nodes, it will allocate 2GB from each NUMA node for the instance in question. So be careful, if you are using CPU affinity.  As even using CPU affinity, lets say you only want to use NUMA 1 for an instance; this can lead to foreign memory access. NUMA allocation still happens across all NUMA nodes (at least in SQL Server 2014).  So even though you might only have subset of schedulers online, memory will still be accessed across all NUMA Nodes.  Even though with new hardware foreign memory access does not have major impact on performance.  Still recommend avoiding it.  If you must use CPU affinity, then allocate sub-set of the processor from each NUMA node.

Lazy Writer Responsibility

Lazy Writer is back ground process and a thread is created per physical NUMA node (for Soft-NUMA, we only get one thread, you can read details here by Jonathan Kehayias (Blog|Twitter)). Lazy Writer responsibility is to make sure sufficient number of free pages, about 640 pages, are always available for new data to be loaded.  If Lazy Writer thread is spinning, that is usually a clear signal of internal memory pressure.

Note: Internal Memory pressure does not mean server doesn’t have enough memory.  There are multiple factors that can contribute to this, one fact is low Max Server Memory, which is what has been explored in these articles.

Original Scenario: SQL Server 2012 Latest Build + Windows 2008 R2 Latest Build + Locked Pages In Memory + Low Max Server Memory + AMD Processors + NUMA Configuration

In this build, I can consistently reproduce the issue.  If I set the Max Memory low enough, it caused the Lazy Writer on NUMA node 0 to spin non-stop, as I documented in SQL Server 2012 NUMA Node Imbalance – Cont’d.  To further to this I understand some additional internals around this thanks to colleague who helped here.  As I was not true sure what is happening.

If we looked at the DBCC MEMORYSTATUS, in particular we notice following for Node 0:

Memory node Id = 0                       KB
—————————————- ———–
VM Reserved                              127086340
VM Committed                             642452
Locked Pages Allocated                   55996
Pages Allocated                          54008
Pages Free                               0
Target Committed                         512000
Current Committed                        698448
Foreign Committed                        0
Away Committed                           0
Taken Away Committed                     0

Review the two numbers high-lighted above, Target memory is what this NUMA node is allocated and Current is what is NUMA consuming, therefore the Page Free count is 0.  Thus Lazy Writer spinning to free up memory.  However it is unable, thus spinning indefinitely causing CPU to pin.  Question to answer, why can’t Lazy Writer free up memory from NUMA Node 0?

Looking further into MEMORYSTATUS, we also notice, that most of the memory allocated into NUMA Node 0 is for buffer cache:

MEMORYCLERK_SQLBUFFERPOOL (node 0)       KB
—————————————- ———–
VM Reserved                              103971820
VM Committed                             524288
Locked Pages Allocated                   2524
SM Reserved                              0
SM Committed                             0
Pages Allocated                          8

Test Scenario #1: Windows 10 + SQL Server 2012 RTM + Low Max Server Memory + Intel Processors + No-NUMA (Physical)

In this, I was just trying to reproduce the issue on my desktop, with no luck.  Memory allocation or such were nominal.

Test Scenario #2: Windows 2012 + SQL Server 2014 RTM + Low Max Server Memory + Intel Processors + No-NUMA (Hyper-V)

Again no issue, as per the title of the series, NUMA Imbalance.  It was obvious and expected this two scenarios don’t show any issues.  But I wanted to confirm it.

Test Scenario #3: SQL Server 2008 R2 + Windows 2008 R2 + Low Max Server Memory + AMD Processors

These test was on same servers as original scenario, however we did not notice any issue.

*** During testing, we realized an interesting fact about NUMA configuration in SQL Server 2008 R2.  SQL Server swaps NUMA 0 and NUMA 1 CPU mapping when creating a logical mapping for schedulers.  For example, if you have 2 NUMA Nodes each with 4 cores; SQL Server binds NUMA 0 – Processor 0 – 3 to Scheduler 4 – 7 and NUMA 1 – Processor 4 – 7 to Scheduler 0 – 3.  This is considered NUMA Swap, documented in this MSDN blog article here by CSS Team.  However this behavior does not seem to persist in SQL Server 2012.

Test Scenario #4: SQL Server 2014 + Windows 2012 R2 + Low Max Server Memory + Intel Processors + NUMA (Azure IaaS Virtual Machine)

This was interesting case, when I set the Max Server Memory to low value, the NUMA node 0’s processor time started to increase.  However it was not as swear as SQL Server 2012, however these are different processors.  I had same behavior as noticed above, Current Committed was higher then Target Committed, therefore NUMA thread on Node 0 was spinning.  However interesting thing note, it was not as aggressive as SQL Server 2012.  It was not continuously spinning, it allowed CPU to breath.  However would wake up frequently as low as ever 20ms.

!!! Eliminated Locked Pages in Memory as a factor for NUMA Imbalance.

Test Scenario #5: SQL Server 2014 + Windows 2012 R2 + Low Max Server Memory + Intel Processors + Locked Pages In Memory + NUMA (Azure IaaS Virtual Machine)

Exact same behavior as above was noticed, however, after restart, SQL Server no longer had Current Committed higher then Target Committed.  Even after lowering the max memory even further then Scenario #4.  It did not cause as dramatic shift.

*** Another interesting thing to note in SQL Server 2014. Up-to SQL Server 2012, Lazy Writer Thread bound to CPU 0 of each NUMA Node.  In SQL Server 2014, Lazy Writer Thread is binding to last CPU of each NUMA Node.

Conclusion & Things to Consider

  • Configuring SQL Server Max Server Memory low memory will cause Lazy Writer to spin due to internal memory pressure issue.
  • There is no value at which point this behavior starts, it depends on the load of the server.  For example, in Original Scenario, due to server load under 4000MB it started spinning the Lazy Writer thread; however in Test Scenario #4, I had to go down to as much as 1000MB. Therefore on your servers look at Lazy Writer thread in conjunction with CPU and DBCC MEMORYSTATUS to understand what is truly happening.
  • This behavior is consistent in whether I have Locked Pages in Memory or not.
  • In Windows 2012 R2 NUMA allocation seem to be much better then Windows 2008 R2.
  • SQL Server 2014 Lazy Writer thread does not seem to be as aggressive as SQL Server 2012.
  • Few KB articles came up during my research in SQL Server 2012 that fix NUMA related issues, KB2819662 and KB2926223.
  • There as been some fixes released for NUMA for Windows 2008 R2 operating system also, please review KB2155311 and KB2510206.
  • In addition, if you are running a server with more then 64-logical cores, please review K-Group Configuration, for details please reference this article.
  • Last but not least verify that memory is distributed evenly on physical server.  That is the memory banks have been evenly allocated, i.e. you do not have more memory allocated in one bank versus another, as you might also see one NUMA node working extra harder if there is significant amount of foreign memory access.

Outstanding Question

Question to answer, why can’t Lazy Writer free up memory from NUMA Node 0?

SQL Server Lazy Writer is only responsible for clearing memory from Buffer Pool, it cannot remove or clear memory from any other memory clerk.  Therefore, it is possible, that other internal components have memory allocated on NUMA Node 0 only, therefore NUMA Node 0 is unable to free up enough memory for buffer pool only.  Looking at DBCC MEMORYSTATUS we can see which memory clerks allocate from NODE Node 0 only, there are few, for example, MEMORYCLERK_SQLCLR, MEMORYCLERK_SOSMEMMANAGER, OBJECTSTORE_LOCK_MANAGER, and MEMORYCLERK_XE_BUFFER all have allocation from NODE Node 0 only.

This post is cross posted on my SQLCAN Blog, MSDN Blog, and SQL server Consultation blog.

Advertisements

SQL Server 2008 SP4 and SQL Server 2008 R2 SP3 Released!

September 29, 2014 2 comments

SQL Server 2008 SP4 (link) and SQL Server 2008 R2 SP3 (link) have been released.  Please note these are the final service packs for these products is now in Extended Support cycle of Product Support Life Cycles.

Update Sept. 30, 2014: Sorry for mistake, for SQL Server 2008 it would be SP4.

Online Index Rebuild can Cause Fragmentation

June 9, 2014 Leave a comment

As Microsoft SQL Server PFE, I am always playing with SQL Server.  I generally spend most of my time on the engine side of world; so I am somewhat of a newbie when it comes to BI.  So I understand when I see something in BI world and am a bit lost.  However when I ran into this, rebuilding index online caused fragmentation.  I was some what baffled. You can try it if you wish, I tested it on SQL 2008, SQL 2008 R2, and SQL 2012.

-- Create table for testing.
CREATE TABLE Test (Col1 INT IDENTITY(1,1), Col2 CHAR(50), Col3 CHAR(50), Col4 Char(50)
CONSTRAINT pkTest PRIMARY KEY
(
        [Col1]
) WITH (ALLOW_PAGE_LOCKS  = OFF)
)

-- Insert test data.
INSERT INTO dbo.Test (Col2, Col3, Col4)
SELECT TOP 1000000 REPLICATE('A',50),REPLICATE('A',50),REPLICATE('A',50)
FROM sys.all_objects O1
CROSS JOIN sys.all_objects O2

-- Check fragmentation.
SELECT * FROM sys.dm_db_index_physical_stats(db_id(),null,null,null,null)

-- Rebuild Index online using MAXDOP
ALTER INDEX pkTest ON Test
REBUILD
WITH (Online=ON, MAXDOP=8)

-- Check fragmentation.
SELECT * FROM sys.dm_db_index_physical_stats(db_id(),null,null,null,null)

Some of you might noticed an interesting property.  ALLOW_PAGE_LOCK.

So turns out this is by design.  If ALLOW_PAGE_LOCK is turned off, we are building a index using parallelism, and building it online; it can cause fragmentation (Ref 1 & Ref 2).

Well lesson learned for future and no, it will not be last one.

SQL Server 2008 R2 SP2 CU6 Released!

April 18, 2013 Leave a comment

SQL Build: 10.50.4279 KB: 2830140

Many fixes are same as those released for SP1 CU12.  But there are some additional fixes in this CU for PowerPivot, SSRS, Cursors, Encryption, one for Database Mirroring.

SQL Server 2008 R2 SP1 CU12 Released!

April 17, 2013 Leave a comment

Build: 10.50.2874 KB: 2828727

Mostly engine fixes, few fixes for SSAS.  Some interesting fixes around SQL Server freezing on shutdown, Assertion check failed/access volition for partitioning, index size issues with RCSI, data corruption.

SQL 2008 R2 SP1 CU10 Released!

January 10, 2013 Leave a comment

Build: 10.50.2868 KB: 2783135

A few fixes have been made, however for some reason I think I have seem most of this in previous CUs.   Still fixes for the memory leak with AUTO_UPDATE_STATISTICS_ASYNC, non-yeilding schedulers for LOB objects, etc. fixes are included.  So if yo have seen similar issues might be worth looking it.

SQL Server 2008 R2 SP2 CU4 Released!

January 9, 2013 Leave a comment

Build: 10.50.4270, KB: 2777358

Many fixes are rolled up in this KB, some of the interesting ones for me (Since I primary focus on the engine) are..

KB2708012, EventID 833 logged in error log when VSS snapshot backup is taking place and write operating occurs.

This is interesting, because Error 833 is ..

SQL Server has encountered <n> occurrence(s) of I/O requests taking longer than 15 seconds to complete on file [<full path of file>] in database [<database name.>] (<dbid>). The OS file handle is <file handle>. The offset of the latest long I/O is: <offset in hexadecimal>

This is usually you have disk I/O issues, but in this case nope. It is a “internal members of the overlapped structure is not initialized correctly”.

KB2770258, You receive in incorrect result when you run a query against a partitioned table in SQL Server 2008 R2.

There are other issues that have been fixes in this CU, might be worth a quick gander to see if any of these apply to you.

 

 

 

Microsoft Support policy for SQL Server in Windows 8 and Windows 2012

November 9, 2012 Leave a comment

Now that Windows 8 (which btw ROCKS!) and Windows 2012 is out, Microsoft has released a KB highlighting the support for SQL Server in respect to these operating systems.

In KB2681562, lists the support requirements, issues with install, etc. So a one-stop-shop or so… 🙂

rReplication Configuration Issue

November 20, 2011 1 comment

Client Question

Replication setup keeps failing, permissions are correct.  Why does it keep giving me error code 5 (access denied).

Error Description

Command attempted:

\\ServerName\R$\Microsoft SQL Server\MSSQL10_50.MSSQLSERVER\MSSQL\repldata\unc\ReplFolder\WOI_WorkOrde180608a5_10125.pre
(Transaction sequence number: 0x0000005F0001433C00CC00000000, Command ID: 10128)

Error messages:

The process could not read file ‘\\ServerName\R$\Microsoft SQL Server\MSSQL10_50.MSSQLSERVER\MSSQL\repldata\unc\ReplFolder\20111114154553\WOI_WorkOrde180608a5_10125.pre’ due to OS error 5. (Source: MSSQL_REPL, Error number: MSSQL_REPL21022)
Get help: http://help/MSSQL_REPL21022

Access is denied.
(Source: MSSQL_REPL, Error number: MSSQL_REPL5)
Get help: http://help/MSSQL_REPL5

The error message is fairly clear; OS Error 5 is error messages returned from Windows API when it is trying to read from the file.  But why was he getting this error?  All permissions are set; SQL Server (Publisher) is able to write to it, why wouldn’t the SQL Server (Subscriber) be able to read it if it is running under same service account?

@SQLCAN: Replication Configuration Issue - Setup Screen

Replication Configuration Issue – Setup Screen

If you take note of the image the location the file is written to; its \\ServerName\R$\Microsoft SQL Server\MSSQL10_50.MSSQLSERVER\..; problem with this is the R$ share.  This is an administrative share, you must grant the SQL Server service account local administrative permissions on your Distributor to be able to read the file from there; which is against security best-practices.  Therefore a better way to do is to use UNC paths.

  1. Create a new folder on you Distribution Server.
  2. Grant your SQL Server Service Account(s), if you are using more than one, permissions to read/write to the folder.
  3. Create a HIDDEN share on the folder (any share with $ after words is hidden share; I generally think anything that doesn’t need to be public on network; hide it).
  4. Control the permission on the share also; so only SQL Server Service Account can read/write to it.
  5. Re-setup your Replication using \\ServerName\ShareName$\.

Client confirmed it worked for them 🙂

%d bloggers like this: